Electrical Engineering and Computer Science Technical Seminar Series
"Scale- and Context-aware Convolutional Neural Networks for Non-Intrusive Load Monitoring"
Dr. Yu Zhang, University of California, Santa Cruz
Faculty Host: Wan Du
Friday, September 20, 2019
12:00 p.m. in COB 263
Abstract: Non-intrusive load monitoring (NILM) refers to the challenging task of disaggregating the power consumption of a household into appliance-level electricity usages. By detecting load malfunction and recommending energy reduction programs, cost-effective NILM provides intelligent demand side management for both utility companies and end users. In this talk, we will introduce a novel neural network architecture for NILM named scale- and context-aware network (SCANet), which exploits multi-scale features and contextual information. Specifically, we develop a multi-branch architecture with multiple receptive field sizes and branch-wise gating. A self-attention module is built to facilitate the integration of global context. Extensive numerical tests on open datasets show that the proposed SCANet significantly outperforms state-of-the-art methods. We will also demonstrate the working mechanisms of the modules by visualizing the network’s intermediate layers.